Title:Adaptive Queue Prediction Algorithm for an Edge Centric Cyber Physical System Platform in a Connected Vehicle Environment

Abstract:In the early days of connected vehicles (CVs), data will be collected only from a limited number of CVs (i.e., low CV penetration rate) and not from other vehicles (i.e., non-connected vehicles). Moreover, the data loss rate in the wireless CV environment contributes to the unavailability of data from the limited number of CVs. Thus, it is very challenging to predict traffic behavior, which changes dynamically over time, with the limited CV data. The primary objective of this study was to develop an adaptive queue prediction algorithm to predict real-time queue status in the CV environment in an edge-centric cyber-physical system (CPS), which is a relatively new CPS concept. The adaptive queue prediction algorithm was developed using a machine learning algorithm with a real-time feedback system. The algorithm was evaluated using SUMO (i.e., Simulation of Urban Mobility) and ns3 (Network Simulator 3) simulation platforms to illustrate the efficacy of the algorithm on a roadway network in Clemson, South Carolina, USA. The performance of the adaptive queue prediction application was measured in terms of queue detection accuracy with varying CV penetration levels and data loss rates. The analyses revealed that the adaptive queue prediction algorithm with feedback system outperforms without feedback system algorithm.